Available at

4.1.2 Lexical density 
Halliday (1985: 62) says that lexical density is “the number of lexical items, as a 
proportion of the number of running words” and adds that “lexical density if the 
kind of complexity that is typical of written language”. Johansson (2008: 61) 
defines lexical density as “a measure of the proportion of lexical items (i.e. nouns, 
verbs, adjectives and some adverbs) in the text”. Lower lexical density suggests 
that the text is less informative. Information is diluted with more function words. 
Between two texts of the same length, the one containing more lexical words 
would be considered more informative.
Before calculating the lexical density of all speeches, it is important to define 
lexical and grammatical words. There is first a distinction to be made between 
closed-class and open-class parts of speech (Jurafsky & Martin, 2004: 3): 
“Closed classes are those that have relatively fixed membership. For example, 
prepositions are a closed class because there is a fixed set of them in English; 
new prepositions are rarely coined. By contrast nouns and verbs are open classes 
because new nouns and verbs are continually coined or borrowed from other 
languages”. Closed-class words are grammatical words, whereas open-class 
words are lexical words. Grammatical words mainly comprise prepositions, 
determiners, pronouns, auxiliaries, conjunctions, particles, numerals, interjections, 
negatives, greetings, and politeness markers. The main groups of lexical words 
are nouns, adjectives, verbs (whether lexical or grammatical) and adverbs.

---

Results and discussion 
 
page 60 
Laviosa (1998) worked on lexical density in TEC, the Translational English Corpus. 
This corpus is made of translated narrative prose (from a number of languages 
into English) and of original English narrative texts. She found that translated texts 
have a significantly lower number of content words versus grammatical words, and 
their lexical density is thus lower (1998: 563). Sandrelli & Bendazzoli investigated 
EPIC to see if this was also accurate for interpreted speeches. The effect noted by 
Laviosa (1998) was not confirmed in their study as the variation between 
interpreted and original speeches is very little. Kajzer-Wietrzny (2005) also 
showed that there was no consistent difference in the informativeness of 
interpreted and original speeches, as lexical densities were very similar (57% for 
original English and 56% for interpreted French). 
As it is shown in Appendix 3, I have distinguished between lexical, grammatical 
and irrelevant POS-tags. Three POS-tags were not considered in the lexical 
density scores reported in Table 10: LS (list markers), SENT (sentence-break 
punctuation) and SYM (symbol) because they were irrelevant to the calculation of 
lexical density. 
I will first take a look at the lexical density of the four source speeches. Table 10 
shows that all source speeches are rather dense (between 52% and 58%). The 
lexical density scores are higher than those obtained in spoken corpora. Ure (1971) 
argues that most spoken texts have a lexical density of under 40%, while a large 
majority of written texts have a lexical density of 40% or higher. This seems to 
suggest that the source speeches included in LOCOSSI are closer to writing than 
speech. 

---

Results and discussion 
 
page 61 
Table 10 – IN lexical density 
IN 
Number of words Number of lexical words 
Lexical density 
IN01 
1,378 
772 
56.02% 
IN02 
1,343 
718 
53.46% 
IN03 
1,487 
860 
57.83% 
IN04 
1,414 
740 
52.33%